xkcd is an enjoyable web cartoon, often with a scientific bent. This time they got it wrong, though.
Introspection is not a proper research method to generate hypotheses. Rather you generate hypotheses from a theory that you are trying to validate. The development of that theory in turn is (should be) subject to the use of proper theory-building research methods like qualitative surveys or case study research.
It is the hallmark of a good theory that it can generate testable hypotheses, which are effectively predictions about the future. If the hypothesis test fails, soemthing is wrong with the theory. This way, a hypothesis is an outgrowth of something larger (the theory) and the testing of the hypothesis contributes to the (in)validation of the theory.